EP1929691B1

EP1929691B1 - Resource allocation method for MIMO-OFDM of multi-user access systems

Info

Publication number: EP1929691B1
Application number: EP05792221A
Authority: EP
Inventors: Peigang Jiang; Mattias Wennstrom
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2005-09-30
Filing date: 2005-09-30
Publication date: 2012-03-14
Anticipated expiration: 2025-09-30
Also published as: US20080165883A1; PT1929691E; EP1929691A1; EP1929691A4; US8238462B2; CN101218775B; CN101218775A; WO2007036073A1; ATE549802T1

Abstract

A resource allocation method for MIMO (multi-input multi-output-OFDM (orthogonal frequency-division multiplex) of multi-user access systems includes A) for each sub-carrier or group of sub-carriers of OFDM, grouping signature vectors of users at a time period according to correlations of the signature vectors; B) from the grouping results, selecting the signature vectors according to a scheduling rule; assigning the sub-carrier frequency and time resource to users of which simultaneously the selected signature vectors have low correlations; and assigning spatial resource to the users corresponding to the selected signature vectors. By minimizing the spatial co-channel interference to an acceptable low level, the complexity significantly in the joint multi-user optimization is reduced.

Description

Field of the invention

The invention relates to resource allocation in multi-user wireless communication system with multiple antennas at both the serving base station and the receiving mobile station, and more specifically to a method for allocating resource in frequency, time, and space dimension with reduced complexity for joint multi-user optimization in MIMO (multiple-input multiple-output) systems with OFDM (orthogonal frequency- division multiplexing) signaling, i.e. MIMO-OFDM systems.

Background of the invention

In wireless MIMO communication systems, the use of multiple antennas is much preferred in order to increase the performance of the systems. When multiple antennas at the base station and mobile station are used, the space dimension may additionally be exploited for scheduling the transmission to different users in the systems.
It is well known that in the multi-user MIMO system, there exit parallel channels characterized by mutually orthogonal spatial signature vectors (SSVs) for each user, data pre-coded by these spatial signature vectors can be transmitted simultaneously without co-channel interference (CCI). If the spatial signature vectors of different users are mutually orthogonal, the data of the simultaneous users can be multiplexed in space domain. To improve the link performance of an MIMO system, knowledge of the MIMO channels should be utilized and pre-coding at the transmitter should be performed. A common well-known pre-coding method is to obtain the pre-coding vectors from the singular value decomposition (SVD) of the MIMO channel matrix. Each singular vector form the MIMO channel matrix, also known as a spatial signature vector, corresponds to a spatial data-stream. In accordance with this approach together with a water filling algorithm to distribute the transmit power among the data streams, the channel capacity for the point-to-point Gaussian channel of single user is achieved.
An OFDM system allows for scheduling of data to users in the time-frequency domain. By using multiple antennas at the base station and the mobile station, users can be scheduled additionally in the spatial domain. When an OFDM time-frequency resource is reused in the spatial domain, this denotes that multiple data-streams are transmitted. Due to the non-orthogonality between spatial data-streams, the data transmitted in different spatial streams for a particular OFDM time-frequency resource, mutually interferes with each other, creating co-channel interference (CCI). Specifically, when transmitting to several users simultaneously in different spatial data-streams of a time-frequency resource in a MIMO-OFDM system, these different users also experiences CCI between the multiple data-streams.
Therefore, in contrast to the time-frequency grid in OFDM, which gives orthogonal channels between different users, the further division of the available channels in the space dimension generates a set of channels that generally are not orthogonal to each other. So, for instance, if a time-frequency resource in an OFDM system is allocated to two users at the same time, the transmission to these users mutually interferes with each other due to the co-channel interference.
Figure 1 shows a transmitting and receiving structure of MIMO-OFDM system. With the multi-user MIMO-OFDM system, the task of the scheduler is to assign OFDM time-frequency resources, as well as spatial resources to the served users in the coverage of the cell. This task can be overwhelming. The problem is that the optimal resource allocation problem in power, space, time and frequency dimensions has to be done jointly among all the users in the coverage of the cell, due to the non-orthogonality of the spatial dimension that couples the signal to interference ratio of the different users through the co-channel interference. If the MIMO channels hypothetically were orthogonal also in the spatial domain, the scheduler's task would be greatly simplified because it could treat the space, time and frequency domain as orthogonal resources to be allocated to the users.
The non-orthogonality in the spatial domain leads to an extremely complex optimization problem. In a first prior art, the resolution titled with "MIMO for Long Term Evolution", which is presented in 3GPP TSG RAN WG1 R1 - 050889, relates to a method for resource allocation. Based on unitary matrix pre-coding, data streams from different users are multiplexed in space and time domain. A set of unitary pre-coding matrices is defined off-line. Major steps in the above prior art are:
1. Each user feedbacks a preferred unitary pre-coding matrix whose column vectors are orthogonal spatial signature vectors. In addition, Channel Quality Information (CQI) for all the pre-coding vectors in the matrix is fed back to the base station;
2. The base station groups users who declare the same preferred unitary pre-coding matrix;
3. The base station selects a group according to a scheduling rule;
4. The base station allocates the orthogonal corresponding pre-coding vectors to users in this selected group according to a scheduling rule;
5. The base station pre-codes different users' data streams by the assigned pre-coding vectors and transmits the pre-coded data streams at the same frequency and time resource.
There are some drawbacks in scheme of the first prior art:
1) The users are always asked to feedback all available spatial signature vectors and corresponding CQI information. In practice, usually only a few or one of the spatial signature vectors and a few or one of CQI information is necessary in the base station.
2) In this scheme, because users are first grouped according to the same pre-coding matrix, this means that only a subset of the users which select the same pre-coding matrix is possible to be multiplexed in space domain. Although this method can guarantee spatial signature vectors of multiplexed user to be strict orthogonal, leading to low CCI, in practice, the group of users who can be spatially multiplexed are restricted. Because the probability that more than one user prefers the same pre-coding matrix, which also is a requirement for spatial multiplexing, may be low in some environment especially when the number of users in the cell coverage area is small, which brings out a serious limitation.
3) There exits such a case that users which prefer different pre-coding matrices and have mutually orthogonal space signature vectors can be multiplexed in space with low CCI, however in the scheme of the prior art this space domain multi-user multiplexing opportunity is lost.
4) The prior art assumes that the selected pre-coding vectors are orthogonal. However, it is easily shown as follows, that this does not assure that the received signals are CCI free.
Assume two users with MIMO channel matrices H ₁ and H ₂ respectively. The data x ₁, to user 1 with the pre-coding vector w ₁ and the data x ₂ to user 2 with the pre-coding vector w ₂ are transmitted. The received signals for user 1 and 2 can then be expressed as $y_{1} = H_{1} (w_{1} x_{1} + w_{2} x_{2}) + n_{1}$
$y_{2} = H_{2} (w_{1} x_{1} + w_{2} x_{2}) + n_{2}$
where n ₁ and n ₂ are the respective noise vectors for the two users. The CCI for user 1 is given by the term H ₁ w ₂ x ₂. In the prior art, it is assumed that pre-coding vectors are orthogonal, hence $w_{1}^{*} w_{2} = 0$
is held. To remove the CCI for user 1 completely, it is necessary that w ₁ is a singular vector to H ₁ and to remove the CCI for user 2 completely, it is necessary that w ₁ is a singular vector to H ₁. Simultaneously $w_{1}^{*} w_{2} = 0$
shall be hold. Hence, complete removal of CCI between two users is very unlikely because it requires a very special relation between the singular vectors of their MIMO channels H ₁ and H ₂.
Therefore, a problem in the prior art lies that it is such a premises based on the requirement of orthogonal pre-coding vectors $w_{i}^{*} w_{j} = 0, i \neq j,$
and such a condition that in reality has low significance because CCI in all practical MIMO channels is non-zero anyway.
To increase the opportunities of multiplexing users, especially when there are a low number of MIMO users in the cell coverage area, to obtain an increase in the multi-user diversity gain, the CCI between users multiplexed in space to be larger than zero is allowed. In other words, if the CCI between users multiplexed in space dimension is allowed to be larger than zero, the opportunity of multiplexing users increases and more diversity gain may be obtained in space domain. Thus both more diversity gain obtained in space domain and low CCI exists conflict and the scheme has no ability to solve the conflict between CCI and multi-users diversity gain or to compromise it.
In a second prior art, an article tilted with "An Efficient Resource-Allocation Scheme for Spatial Multi-user Access in MIMO/OFDM Systems", is published in IEEE Transactions, on Communications., Vol.53, No.1, 1, 2005. This article discloses that radio resources are exploited in frequency, time and space domain. Users are grouped according to user's mutual correlations which respectively depend on the maximal SSV selected from spatial signature vectors of each user. The scheduler is potential to assign the radio resources with the same frequency and time to different users if correlations between any pair of users from different groups are sufficient low, i.e. the users are orthogonal in space domain.
However, the resources allocation scheme is suitable for uplink and only the space mode with maximum gain is considered for each user. In practical wireless communication environments, the grouping criterion may not always be fulfilled. Furthermore, the amount of required feedback signalling is large.
Another prior art is that of the document "Comparison of Zero-forcing Methods for Downlink Spatial Multiplexing in Realistic Multi-user MIMO Channels" by Giovanni Del Gado and Martin Haardt, in which spatial signature vectors of users are grouped according to their mutual correlation. The same time resources are assigned to each of a group of users for which the spatial signature vectors have low mutual correlations.

Summary of the invention

This invention presents a method for allocating resource in frequency, time, and space dimension to reduce the complexity significantly in the joint multi-user optimization.
Instead of grouping users having all the same SSVs in the first prior art, the SSVs of each user are independently grouped with other user's SSVs. That means multiplexed users need not have the same preferred pre-coding matrix as in the first prior art, also the grouping is controlled by a parameter of SSV correlation which determines the CCI. Choosing different parameters, different compromising results between CCI and multiplexing opportunity in space domain can be obtained. The throughput and Quality of Services (QoS) of the system can be enhanced by optimizing the parameter. Instead of grouping users in the second prior art, it is easy to guarantee orthogonality of the spatial dimension with users' SSVs directly of the invention. Thus, by minimizing the spatial co-channel interference to an acceptable low level, the complexity significantly in the joint multi-user optimization is reduced.
Different from the first prior art and the second prior art, the number of SSVs fed back from each user is adaptive and individual for each user depending on the particular user's channel condition such as channel rank.
In summary, effects of this invention could be described as:
The invention improves over the prior art by firstly removing the pre-coding matrix design criterion in prior art that pre-coding vectors are orthogonal, thereby, a better match between pre-coding vectors and channel singular vectors.
Secondly, the grouping in the spatial domain is performed among the users individually selected (i.e. not group selected) preferred pre-coding vectors instead as between users as is done in the prior art.
Thirdly, the flexibility is further increased by allowing an adaptive number of pre-coding vectors that are fed back according to the users channel conditions.
Fourthly, a feedback method is described which allows the grouping to be made off-line, prior to real time operation. These improvements described in this disclosure give the time-frequency-space resource allocation scheduler in a MIMO/OFDM system.

Brief Description of the Drawings

Figure 1 shows a transmitting and receiving structure of MIMO-OFDM system;
Figure 2 shows the flowchart of the radio resources allocation method;
Figure 3 shows the grouping steps according to a rule of un-correlation within group; and
Figure 4 shows the grouping steps according to a rule of un-correlation between groups.

Embodiments of the Invention

The present invention is described in detail in conjunction with the accompanying drawings. A resource allocation for downlink from the base station to users is taken as an embodiment.
The SSVs and the spatial sub-channel gains, SSGs, for each OFDM time-frequency resource from the multiple users are received by a base station. In the base station, these SSVs are grouped in several groups where the SSVs within a group have high spatial correlation and SSVs from different groups have low spatial correlation, or the SSVs within a group have low spatial correlation and SSVs from different groups have high spatial correlation. The groups are defined by specifying a spatial correlation threshold parameter.
Based on the first grouping rule where the SSVs within a group have high spatial correlation and SSVs from different groups have low spatial correlation, the base station selects one SSV and corresponding user from each group according to a scheduling rule. Data streams from selected users are pre-coded by associated SSV and transmitted. Based on the second grouping rule where the SSVs within a group have low spatial correlation and SSVs from different groups have high spatial correlation, the base station selects one group according to a scheduling criteriion. Data streams are pre-coded by SSVs in this group and transmitted.
To reduce feedback load, each user selects its SSV from a pre-determined set of SSVs and feeds back only the indices of the table with pre-determined set of SSVs and the corresponding SSGs. The table is stored in both the base station and the users. This reduces the feedback load considerably. Because there is a predetermined set of SSVs, the grouping of SSVs in the first part of the invention can be made off-line. Hence, when an SSV index is received from a user, it may be directly categorized in one of the pre-determined groups. This reduces the complexity of the present invention considerably.
With this separation of the SSVs in groups, the complexity of the optimisation in the joint space-time-frequency resource allocation is reduced because the CCI in the space domain is below a pre-determined value and the space-time-frequency resources may be considered close to orthogonal, so that the CCI entanglement in the joint multi-user optimisation is decoupled.
A multi-user access MIMO-OFDM system is modeled as follows: $y_{k}^{l} (t) = H_{k}^{l} (t) x (t) + n_{k}^{l} (t), k = 1, \dots, K; l = 1, \dots, L,$
where $H_{k}^{l} (t) \in C^{m \times n}$
is wireless channel coefficient matrix from the base station to kth user corresponding to lth sub-carrier at sampled time t, $y_{k}^{l} (t) \in C^{m}$
is the received vector of kth user on lth sub-carrier at sampled time t, x(t) ∈ C ⁿ is the transmitted symbol vector satisfying average power constraint E,{x ^H (t)x(t)}≤ P, $n_{k}^{l} (t) \in C^{m}$
is the corresponding additive channel noise. Because the resources are orthogonal in time-frequency domain, in the following the time index t and frequency superscript l without confusion are omitted.
Any matrix, especially the channel coefficient matrix for each user k can be decomposed by SVD as follows:
where $\begin{matrix} H_{k} = U_{k} D_{k} V_{k}^{H} \\ U_{k} = [\begin{matrix} u_{k, 1} & u_{k, 2} & \dots & u_{k, m} \end{matrix}] \in C^{m \times m} \\ V_{k} = [\begin{matrix} v_{k, 1} & v_{k, 2} & \dots & v_{k, n} \end{matrix}] \in C^{n \times n} \end{matrix}$

are unitary matrices and D ∈ C ^m×n is a diagonal matrix. If rank(H _k )= ω _k , the SVD of the MIMO channel matrix provides ω _k parallel (no mutual interference) sub-channels from the transmitter to the receiver. The elements being equal to nonzero in matrix D are the SSGs, and the column vectors of V _k can be regarded as the SSV of each sub-channel. Data streams of kth user weighted by these SSVs can be detected by the receiver without self interference from other SSVs from the same SVD.
When considering multi-user access system, if the spatial signature vectors corresponding to different users are orthogonal, these users can transmit data streams in the same sub-carrier without CCI. For example, if $v_{i, 1}^{H} v_{j, 1} = 0,$
the CCI free transmit vector can be constructed as: $x = \sqrt{P_{1, i}} v_{i, 1} x_{i} + \sqrt{P_{1, j}} v_{j, 1} x_{j}$

where x_i is the data streams of user i, x_j is the data streams of user j and P_l,j is the allocated power of stream l of user j. In practice, the equation $v_{i, 1}^{H} v_{j, 1} = 0$
is highly unlikely, so the CCI of users transmitting data streams in the same sub-carrier and at the same time index t (same time-frequency resource), can be controlled by the correlation coefficient of the spatial signature vectors used to weight the data streams.
Figure 2 shows the steps of the radio resources allocation method. The method includes:
Step 101: For each sub-carrier or group of sub-carriers I and time period t, a base station transmits control information about which users are allowed to transmit data in the time-frequency resource. Preferably in order to decrease channel resource, the control information may be broadcasted to the users. The base station may also broadcast the maximum number of data streams multiplexed in space for each user, for instance, the maximum number is denoted r_k for user k. r_k may be smaller than rank(H _k ) and may be changed adaptively according to the users changing channel condition and the traffic load in the system. In this way, flexibility is improved.
Step 102: The users feedback their spatial signature vectors $\begin{matrix} v_{k, 1}^{l} & v_{k, 2}^{l} & \dots & v_{k, r_{k}}^{l} \end{matrix}$
and the corresponding spatial sub-channel gains $\begin{matrix} λ_{k, 1}^{l} & λ_{k, 2}^{l} & \dots & λ_{k, r_{k}}^{l} \end{matrix}$
to the base station by uplink channel.
Step 103: After having received the SSVs and corresponding SSGs, the base station groups the SSVs for each sub-channel 1 according to a pre-determined threshold ρ. The detailed grouping steps will be described hereafter. The grouping results can be expressed as: $\begin{array}{l} G_{1}^{l} = \{v_{k_{G_{1}, 1, s_{1}}}^{l} v_{k_{G_{1}, 2, s_{2}}}^{l} \dots\} \\ G_{2}^{l} = \{v_{k_{G_{2}, 1, s_{1}}}^{l} v_{k_{G_{2}, 2, s_{2}}}^{l} \dots\} \\ \dots \\ G_{g}^{l} = \{v_{k_{G_{g}, 1, s_{1}}}^{l} v_{k_{G_{g}, 2, s_{2}}}^{l} \dots\} \end{array}$
where $G_{g}^{l}$
represents g^th grouping results for sub-channel l.
Step 104: The base station allocates the spatial, frequency and time resource to the users according to the grouping results. Namely, the base station determines which users can transmit data streams in sub-carrier 1 and time period t, and which SSVs can be used to weight data streams. Different grouping rules lead to different allocation modes.
Step 105: The base station broadcasts the resource allocation results to the users. Also optionally, for enhancing the users' demodulation performance, the SSVs of each data stream can be broadcasted to all users.
Step 106: The base station transmits weighted data streams to corresponding user.
Step 107: The users receive and demodulate the control information and data streams from the base station.
There may be two different grouping methods which are shown in Figure 3 and 4, respectively, used in step 103. Figure 3 shows the grouping steps according to un-correlation within group rule (UWGR). By UWGR, t/he grouping is such that the correlations of SSVs within the same group are lower than or equal to the pre-determined threshold ρ.
The grouping based on UWGR can be implemented at the base station by the following steps:
Step 201: Order the SSVs usually according to a scheduling criterion.
Step 202: For each sub-carrier or group of sub-carriers 1, create a new group $G_{g}^{l}$
and move the first un-grouped SSV to this group.
Step 203: Select one un-grouped SSV in order and check the correlations of the selected SSV with all the SSVs in the group.
Step 204: If all the checked correlations are low than or equal to the pre-determined threshold ρ, move this un-grouped SSV to the new group; otherwise, go to Step 205.
Step 205: Judge whether all the un-grouped SSVs are checked, if yes, go to step 206; otherwise, return to step 203.
Step 206: Judge whether all the received SSVs are grouped, if one of the received SSVs are not grouped, return to step 202; otherwise, to step 104.
Based on the grouping results by UWGR, step 104 is implemented as following: selecting one group according to a scheduling rule, and allocating the frequency and time resource to at least one user whose SSVs are included in this selected group; weighting the data streams of selected user by the corresponding SSV in this group. In this way, the users are orthogonal in space domain so as to decouple CCI.
Figure 4 shows the grouping steps according to un-correlation between groups rule (UBGR). By UBGR, the grouping is such that the correlations of SSVs belonging to different groups are lower than or equal to the pre-determined threshold ρ. The grouping based on UBGR may be implemented at the base station by the following steps:
Step 301: Similar to step 201, order the SSVs according to the scheduling criterion.
Step 302: For each sub-carrier or group of sub-carriers 1, creating a group $G_{g}^{l}$
and moving the first un-grouped SSV to this group.
Step 303; Select one un-grouped SSV in order and check the correlations of the selected SSV with all the SSVs in the group.
Step 304: If all the checked correlations are low than or equal to the pre-determined threshold ρ, create a new group and moving this un-grouped SSV to the this group; otherwise, go to Step 305.
Step 305: Combine the groups where at least one SSV has correlation higher than the pre-determined threshold ρ with the selected SSV.
Step 306: Judging whether all the received SSVs are grouped, if one of the received SSVs are not grouped, return to step 303; otherwise, to step 104.
Based on the grouping results by UBGR, step 104 is implemented as following: selecting one SSV in each group according to a scheduling rule, determining corresponding users according to the selected SSVs so that the corresponding users are selected indirectly, allocating with the frequency and time resource; weighting data streams of the selected user by the corresponding SSV in each group. In this way, the users are orthogonal in space domain so as to decouple CCI.
The groupings given in Figure 3 and Figure 4 are implemented in real time according to users' feedback information, which need a lot of calculation resources of the base station. A practical implementation procedure with less complexity of the method is described as following. First, to reduce the amount of feedback signaling of the SSVs, for each set of n transmitter antennas and m receiver antennas, there is a predetermined set of Mn,m SSVs hereafter called generic SSVs, denoted $S_{n, m} = \{\begin{matrix} w_{1}^{(n, m)} & w_{2}^{(n, m)} & \dots & w_{M_{n, m}}^{(n, m)} \end{matrix}\} .$
These sets of generic SSVs are predetermined in the system and known to both the receiver and the transmitter. When a user wants to feed back an SSV, it searches for the closest generic SSV in the set Sn,m and determines a corresponding index, where "closeness" can be defined in various ways according to the theory of vector quantization.
Instead of feeding back the whole SSV, the user feeds back the SSV index to the "closest" vector wc(n,m) in Sn,m. Hence, only |log₂(M_n,m )| bits are needed for this feedback.
Because by a quantization method all possible SSVs that can be feed back from the users in the systems are fixed and known in advance, the grouping of the predetermined SSVs can also be made in advance to obtain a generic group. Hence, the generic groups can be formed as: $\begin{array}{l} {\tilde{G}}_{1} = \{w_{s_{1}}^{(n_{1}, m_{1})} w_{s_{2}}^{(n_{2}, m_{2})} \dots\} \\ {\tilde{G}}_{2} = \{w_{s_{q}}^{(n_{q}, m_{q})} w_{s_{q + 1}}^{(n_{q + 1}, m_{q + 1})} \dots\} \\ \dots \\ {\tilde{G}}_{g} = \{w_{s_{a}}^{(n_{a}, m_{a})} w_{s_{a + 1}}^{(n_{a + 1}, m_{a + 1})} \dots\} \end{array}$
such that correlations of generic SSVs within each group are lower than or equal to ρ by UWGR. Or alternatively, the correlations of vectors belongs to different groups are lower than or equal to p by UBGR. These generic groups are independent of user index k and sub-band index l.
When the base station receives an SSV index from a user, it may assign SSV corresponding to the index to one of the generic groups without the need for any spatial correlation calculations, thereby grouping results can be obtained with reducing the complexity considerably.
Although illustrative embodiments of the disclosure have been shown and described, other modifications, changes, and substitutions are intended in the above disclosure. For example, other signature vectors used to characterize spatial resource also can be grouped, or the method of invention is used in uplink.

Claims

A resource allocation method for multi-input multi-output - orthogonal frequency-division multiplex, MIMO-OFDM, of multi-user access systems, characterized by comprising:
receiving, by the base station BS, spatial signature vectors, SSVs, fed back from the users, wherein the number of SSVs fed back from each user is adaptive and individual for each user depending on the particular user's channel condition.
A) for each sub-carrier of orthogonal frequency-division multiple x OFDM, grouping, by a base station, BS, the spatial signature vectors at a time period according to correlations of the spatial signature vectors;

B) from the grouping results, selecting, by the base station BS, the spatial signature vectors according to a scheduling rule;
assigning, by the base station BS, the same sub-carrier frequency and time resource to users of which simultaneously the selected spatial signature vectors have low correlations; and
assigning, by the base station BS, spatial resource to the group of users corresponding to the selected spatial signature vectors,
The resource allocation method according to claim 1, wherein the step of grouping signature vectors according to correlations of the signature vectors comprises, according to un-correlation within group rule ,UWGR, getting such grouping results that the correlation of signature vectors within the same group are lower than or equal to a pre-determined threshold.
The resource allocation method according to claim 2, wherein the step of grouping signature vectors comprises,
A11) creating a group and moving the first un-grouped signature vector to this group;

A12) selecting one un-grouped signature vector and checking the correlations of the selected signature vector with all the signature vectors in the group;

A13) judging whether all the checked correlations are low than or equal to the pre-determined threshold, if yes, moving this un-grouped signature vector to the group, otherwise, returning to step A12 until all the un-grouped signature vectors are checked;

A14) returning to step A11 until all signature vectors are grouped.
The resource allocation method according to claim 1, wherein the step of grouping signature vectors according to correlations of the signature vectors comprises, according to an un-correlation between group rule, UBGR, getting such grouping results that the correlation of signature vectors between the different groups are lower than or equal to a pre-determined threshold.
The resource allocation method according to claim 4, wherein the step of grouping signature vectors comprises,
A21) creating a group and moving the first un-grouped signature vector to this group;

A22) selecting one un-grouped signature vector and checking the correlations of the selected signature vector with all the signature vectors in the group; and

A23) judging whether all the cheeked correlations are lower than or equal to the pre-determined threshold, if yes, creating a new group and moving the un-grouped signature vector to this new group, and then returning to step A22 until all the signature vectors are grouped; otherwise, combining the groups where at least one signature vector has correlation higher than the pre-determined threshold with the selected signature vector, and then returning to step A22 until all the signature vectors are grouped.
The resource allocation method according to claim 4 or 5, before the step of creating a group and moving the first un-grouped signature vector to this group, further comprising, ordering all signature vectors according to a scheduling criterion,
wherein the step of selecting one un-grouped signature vector further comprises, selecting one un-grouped signature vector in order.
The resource allocation method according to one of the preceding Claims, the step of assigning spatial resource to the users corresponding to the selected signature vectors comprises, weighting data streams of the users with the users' selected signature vectors.
The resource allocation method according to claim 2 or 3, wherein step B comprises,
selecting one group according to the scheduling rule;
allocating the same sub-carrier or group of sub-carriers to at least one user whose signature vectors are included in this selected group; and
weighting the data streams oaf the user by the corresponding signature vector in this group.
The resource allocation method according to claim 4 or 5, wherein step B comprises,
selecting one signature vector in each group according to the scheduling rule; determining corresponding users according to the selected signature vectors; allocating the same sub-carrier or group of sub-carriers to the determined users; and weighting data streams of the determine users by the corresponding signature vector in each group.
The resource allocation method according to claim 1, further comprising, in advance storing predetermined generic spatial signature vectors SSVs at both a base station and the users;
before step A, further comprising,
broadcasting control information by the base station;
selecting the closest generic SSVs while the users want to feed back spatial signature vectors SSV;
feeding back SSV indices corresponding to the closest generic spatial signature vectors SSVs selected by the users;
step A further comprises,
at a base station, after receiving the SSV indices fed back from the users, searching the generic spatial signature vectors SSVs corresponding to the indices and then grouping the found generic spatial signature vectors SSVs in real-time;
after step B, further comprising,
C) broadcasting the allocated frequency and time resource, and the spatial signature vectors SSV indices used to weight data streams.
The resource allocation method according to claim 1, further comprising, in advance storing predetermined generic spatial signature vectors SSVs at both a base station and the users,
wherein step A further comprises, in advance grouping the predetermined generic spatial signature vectors SSVs to obtain generic groups and storing the generic groups in the base station,
between step A and B, further comprises, assigning spatial signature vectors SSVs corresponding to the received indices to the generic groups to thereby obtain grouping results.
A Base Station in an MIMO-OFDM of multi-user access systems being arranged for resource allocation, characterized in that the base station is configured for the method according to any of claims 1 to 11.